Study Quantifies Two-Sided Receptivity Gap for AI Agents in Online Dating
A recent study posted on arXiv reveals significant differences in user interaction with AI on dating platforms. Titled 'Delegation Asymmetry in Agentic Recommender Systems,' the research assesses two-sided receptivity to AI agents. It analyzed data from nearly 5,500 active individuals, discovering a pronounced preference for users to allow autonomous LLM agents to engage in conversations rather than respond to messages initiated by these agents. With a high correlation coefficient of 0.92 observed, the findings indicate that sending and receiving AI-generated messages are not identical processes, influencing the advancement of AI-driven systems in social matchmaking.
Key facts
- Paper available on arXiv with identifier 2608.18058
- Study based on two surveys of a major dating platform
- Survey on generative profile features had N=2,894
- Survey on autonomous conversational agents had N=2,617
- Surveys were fielded in two languages
- Measurement model used graded response models with latent regression
- Willingness to send and receive agent communication are distinct constructs with rho=0.92 and Delta BIC=52
- Systematic delegation asymmetry found: users prefer sending over receiving agent communication
Entities
Institutions
- arXiv